add Model accuracy alignment - #1751
Conversation
There was a problem hiding this comment.
| 序号 | 位置 | 优先级 | 状态 |
|---|---|---|---|
| 1 | setup_venvs.sh:95 | ✅ | |
| 2 | setup_venvs.sh:141 | ✅ | |
| 3 | Test-release.yml:1009 | ✅ | |
| 4 | Test-release.yml:1004 | ✅ | |
| 5 | Test-release.yml:1402 | ✅ | |
| 6 | setup_venvs.sh:38 | ✅ | |
| 7 | setup_venvs.sh:128 | ✅ | |
| 8 | setup_venvs.sh:34 | ✅ | |
| 9 | setup_venvs.sh:186 | ✅ | |
| 10 | run_paddle_minimax.sh:1 | ✅ | |
| 11 | run_torch_minimax.sh:48 | ✅ | |
| 12 | Test-release.yml:1098 | ✅ | |
| 13 | setup_venvs.sh:29 | ✅ |
环境安装、运行脚本路径、coverage 上传、H20 多卡集成、对齐 workflow、模型路径、版权头及默认 wheel 地址问题均已修复,当前未发现阻塞性问题。
| local torch_py="$1" | ||
|
|
||
| echo "[setup_venvs] torch python : ${torch_py}" | ||
| uv pip install --python "${torch_py}" "torch==${TORCH_VERSION}" |
There was a problem hiding this comment.
已在当前提交将 Torch 安装改为使用 cu130 专用索引,torch==2.12.0+cu130 已可解析,本项已解决。
| "pybind11[global]>=2.13,<3" "paddle-nvidia-nvshmem-cu13>=3.3.9,<3.5" | ||
|
|
||
| # PaddleFleet | ||
| uv pip install --python "${paddle_py}" "${paddle_index[@]}" --no-deps \ |
There was a problem hiding this comment.
There was a problem hiding this comment.
已在当前提交移除 PaddleFleet wheel 安装的 --no-deps,并沿 Paddle cu130 源解析其 paddlepaddle-gpu 依赖,本项已解决。
| export COVERAGE_FILE=/workspace/PaddleFleet/coveragedata-multi/.coverage | ||
| export COVERAGE_RCFILE=/workspace/PaddleFleet/ci/.coveragerc | ||
| bash -x PaddleFormers/tests/integration_test/qwen3vl_sft.sh fsdp h20 | ||
| export PADDLEFLEET_WHEEL_PATH="dist/${{ needs.build_whl.outputs.root_whl_name }}" |
There was a problem hiding this comment.
There was a problem hiding this comment.
已在当前提交将 workflow 传入的两个 wheel 路径改为 /workspace/PaddleFleet/dist/... 绝对路径,与脚本切换目录后的解析位置一致,本项已解决。
| # export COVERAGE_RCFILE=/workspace/PaddleFleet/ci/.coveragerc | ||
| # bash -x PaddleFormers/tests/integration_test/qwen3vl_sft.sh fsdp h20 | ||
| # ' | ||
| - name: alignment_model_accuracy |
There was a problem hiding this comment.
当前提交已将 H20 multi-card job 及其 coverage 上传步骤整体注释,并从 coverage_check.needs 中移除该 job;当前 workflow 不再执行无产出目录的上传步骤,本项已解决。
| echo -e "\033[32mIntegration test succeeded: GLM4.5.\033[0m" | ||
| fi | ||
| ' | ||
| # - name: GLM4.5 pre-train |
There was a problem hiding this comment.
当前提交已恢复 H20 multi-card job 及其 GLM/Qwen 覆盖步骤;但 coverage_check.needs(.github/workflows/Test-release.yml:1402)仍未包含 integration-test-H20-multi-card。coverage_check 会下载 h20-multi/coverage.xml,缺少该依赖可能导致它在上传完成前并发执行并读取不到当前产物。请将该 job 加回 needs 后再视为解决。
There was a problem hiding this comment.
已复查当前 HEAD:coverage_check.needs 已包含 integration-test-H20-multi-card,并会在该 job 完成后再下载 h20-multi/coverage.xml。H20 多卡集成步骤及 coverage 依赖均已恢复,此问题已解决。
| export COVERAGE_FILE=/workspace/PaddleFleet/coveragedata-multi/.coverage | ||
| export COVERAGE_RCFILE=/workspace/PaddleFleet/ci/.coveragerc | ||
| bash -x PaddleFormers/tests/integration_test/qwen3vl_sft.sh fsdp h20 | ||
| export PADDLEFLEET_WHEEL_PATH="dist/${{ needs.build_whl.outputs.root_whl_name }}" |
There was a problem hiding this comment.
已在当前提交将 workflow 传入的两个 wheel 路径改为 /workspace/PaddleFleet/dist/... 绝对路径,与脚本切换目录后的解析位置一致,本项已解决。
| readonly UV_BIN_DIR="root/.local/bin" | ||
| readonly UV_CACHE_DIR_PATH="root/.cache/uv" |
There was a problem hiding this comment.
当前提交已为两条路径补上前导 /,分别恢复为 /root/.local/bin 和 /root/.cache/uv,本项已解决。
| echo "[setup_venvs] paddle python: ${paddle_py}" | ||
| local -a paddle_index=( | ||
| --index-url "${PADDLE_INDEX_URL}" | ||
| --index-strategy unsafe-best-match |
There was a problem hiding this comment.
当前提交已将通用构建依赖改回默认 PyPI 索引,同时保留 Paddle wheel 的 stable 源,ninja 和 tensor-spec-worker 不再被限制在 Paddle 源,本项已解决。
| readonly PADDLEFLEET_WHEEL="${PADDLEFLEET_WHEEL_PATH:-${NIGHTLY_WHL_BASE}/paddlefleet/paddlefleet-0.4.0.dev20260807+d01517879a3-py3-none-any.whl}" | ||
| readonly PADDLEFLEET_OPS_WHEEL="${PADDLEFLEET_OPS_WHEEL_PATH:-${NIGHTLY_WHL_BASE}/paddlefleet-ops/paddlefleet_ops-0.4.0.dev20260807+d0151787-cp312-cp312-linux_x86_64.whl}" | ||
| readonly PADDLEFORMERS_WHEEL="${NIGHTLY_WHL_BASE}/paddleformers/paddleformers-0.0.0.dev-py3-none-any.whl" | ||
| readonly MEGATRON_CORE_WHEEL="wget https://paddle-qa.bj.bcebos.com/paddlefleet/whl/megatron_core-0.19.0+f2706b6f3-cp312-cp312-linux_x86_64.whl" |
There was a problem hiding this comment.
当前提交已移除 wget 前缀,MEGATRON_CORE_WHEEL 恢复为可直接传给 uv 的纯 URL,本项已解决。
| ensure_venv "venv/torch" | ||
| ensure_venv "venv/paddle" | ||
|
|
||
| # setup_torch_venv "${WORKSPACE_DIR}/venv/torch/bin/python" |
There was a problem hiding this comment.
当前提交已恢复 setup_torch_venv 调用,并在 Paddle venv 的构建依赖中补上 tensor-spec-worker;Torch/Megatron 对齐环境会在测试前完成安装,本项已解决。
| @@ -0,0 +1,75 @@ | |||
| #!/usr/bin/env bash | |||
There was a problem hiding this comment.
当前提交已恢复两个运行脚本的 Apache 2.0 版权头,本项已解决。
| --model /home/.cache/PaddleFormers/MiniMax-V2.5-bf16_2EP | ||
|
|
||
| ### data | ||
| --dataset /home/.cache/PaddleFormers/MiniMax-V2.5-bf16_2EP/alignment_torch.jsonl |
There was a problem hiding this comment.
当前提交已将 Paddle YAML 的模型和数据路径统一为 /home/.cache/PaddleFormers/MiniMax-V2.5-bf16_2EP,与 Torch 对齐脚本一致,本项已解决。
| echo -e "::error:: \033[31mIntegration test failed: GLM4.5.\033[0m" | ||
| export PADDLEFLEET_WHEEL_PATH="/workspace/PaddleFleet/dist/${{ needs.build_whl.outputs.root_whl_name }}" | ||
| export PADDLEFLEET_OPS_WHEEL_PATH="/workspace/PaddleFleet/dist/${{ needs.build_whl.outputs.ops_whl_name }}" | ||
| model_acc_align_exit_code=$? |
There was a problem hiding this comment.
当前提交已恢复 run_alignment_test.sh 调用,并在执行后紧接着捕获 $?;job 不再空跑,本项已解决。
| readonly TE_VERSION="2.17.1" | ||
| readonly PADDLE_INDEX_URL="https://www.paddlepaddle.org.cn/packages/stable/cu130/" | ||
| readonly NIGHTLY_WHL_BASE="https://paddle-whl.bj.bcebos.com/nightly/cu130" | ||
| readonly FLEET_WHL_BASE="https://paddle-github-action.bj.bcebos.com/whl/" |
There was a problem hiding this comment.
已复查当前代码:FLEET_WHL_BASE 已切换为 Paddle QA 地址,Megatron Core 与 ms-swift 两个默认 wheel 地址均返回 HTTP 200。此前结论基于旧地址,现确认该问题已解决。
PR Category
Execute Infrastructure
PR Types
Improvements
Description
add Model accuracy alignment
是否引起精度变化
否